Diagnostic value of strand-specific miRNA-101-3p and miRNA-101-5p for hepatocellular carcinoma and a bioinformatic analysis of their possible mechanism of action Xia Yang1,†, Yu-Yan Pang1,†, Rong-Quan He2, Peng Lin3, Jie-Mei Cen2, Hong Yang3, Jie Ma2,# and Gang Chen1,#

1 Department of Pathology, First Affiliated Hospital of Guangxi Medical University, Nanning, China 2 Department of Medical Oncology, First Affiliated Hospital of Guangxi Medical University, Nanning, China 3 Department of Ultrasonography, First Affiliated Hospital of Guangxi Medical University, Nanning, China

Keywords There is accumulating evidence that miRNA might serve as potential diag- bioinformatics; Expression Omnibus; nostic and prognostic markers for various types of cancer. Hepatocellular hepatocellular carcinoma; microRNA-101-3p; carcinoma (HCC) is the most common type of malignant lesion but the sig- microRNA-101-5p; The Cancer Genome nificance of miRNAs in HCC remains largely unknown. The present study Atlas aimed to establish the diagnostic value of miR-101-3p/5p in HCC and then Correspondence further investigate the prospective molecular mechanism via a bioinformatic J. Ma, Department of Medical Oncology, analysis. First, the miR-101 expression profiles and parallel clinical parame- First Affiliated Hospital of Guangxi Medical ters from 362 HCC patients and 50 adjacent non-HCC tissue samples were University, Shuangyong Road 6, Nanning, downloaded from The Cancer Genome Atlas (TCGA). Second, we aggre- Guangxi, China gated all miR-101-3p/5p expression profiles collected from published litera- Tel: +86 771 5312980 ture and the Omnibus and TCGA databases. Subsequently, E-mail: [email protected] G. Chen, Department of Pathology, First target of miR-101-3p and miR-101-5p were predicted by using the Affiliated Hospital of Guangxi Medical miRWalk database and then overlapped with the differentially expressed University, Shuangyong Road 6, Nanning, genes of HCC identified by natural language processing. Finally, bioinfor- Guangxi, China matic analyses were conducted with the overlapping genes. The level of miR- Tel: +86 771 5356534 101 was significantly lower in HCC tissues compared with adjacent non- E-mail: [email protected] HCC tissues (P < 0.001), and the area under the curve of the low miR-101

† level for HCC diagnosis was 0.925 (P < 0.001). The pooled summary receiver These authors contributed equally to this work. #These authors contributed equally to this work. operator characteristic (SROC) of miR-101-3p was 0.86, and the combined SROC curve of miR-101-5p was 0.80. Bioinformatic analysis showed that the (Received 19 March 2017, revised 8 May target genes of both miR-101-3p and miR-101-5p are involved in several 2017, accepted 8 November 2017) pathways that are associated with HCC. The hub genes for miR-101-3p and miR-101-5p were also found. Our results suggested that both miR-101-3p doi:10.1002/2211-5463.12349 and miR-101-5p might be potential diagnostic markers in HCC, and that they exert their functions via targeting various prospective genes in the same pathways.

Abbreviations AFP, alpha-fetoprotein; AUC, area under the curve; BP, biological process; CC, cellular component; CI, confidence interval; CNKI, Chinese National Knowledge Infrastructure; DAVID, Database for Annotation, Visualization and Integrated Discovery; FN, false negative; FoxO, forkhead box O; FP, false positive; GEO, Gene Expression Omnibus; GO, ; HBV, ; HCC, hepatocellular carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; MAPK, mitogen-activated kinase; MF, molecular function; NLP, natural language processing; NLR, negative likelihood ratio; PI3K, phosphoinositide-3-kinase; PLR, positive likelihood ratio; PPI, protein– protein interaction; pre-miRNA, precursor miRNA; ROC, receiver operating characteristic; SROC, summary receiver operator characteristic; TCGA, The Cancer Genome Atlas; TN, true negative; TP, true positive.

64 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

According to Cancer Statistics, 2017 [1], the incidence miRNA, they may have different target genes and bio- rates of liver cancer in the USA continue to increase logical functions. A previous study [20] reported that rapidly (~ 3% per year in women and 4% per year in the expression levels of the miR-5p and miR-3p men), and the death rate rose by almost 3% per year mature sequences can be altered in different tissues. from 2010 to 2014. In addition, the mortality rate is Accumulating evidence [19,21–25] has shown that three times higher in men than in women. Since Asia is miR-101-3p/-5p is down-regulated in multiple malig- the area with the highest incidence rate of liver cancer, nances, including HCC. For example, Hou et al. [26] especially China, annual incidence and mortality are explored miRNA expression profiling and revealed more than half of the global totals [2]. Among the three that miR-101 (3p and 5p were not distinguished) histological types of liver malignancy, hepatocellular expression in HCC tissues was lower than in healthy carcinoma (HCC) has become the leading cause of controls. Wei et al. [27] also showed that miR-101 (3p death from cancer. Since there have been no biomarkers and 5p were not distinguished) was down-regulated in or common surgical techniques for the early stage of HBV-associated HCC tissues and may have therapeu- HCC, the majority of patients with HCC are diagnosed tic potential in HCC. Additionally, the function of late, which directly correlates with a poor outcome and these miRNAs has also been investigated. Zhang et al. low survival rate. As with other cancers, HCC develop- [28] revealed that enforced expression of miR-101 (3p ment is a multistep process with abundant genetic and and 5p were not distinguished) by siRNA inhibited the epigenetic mutations. A recent study confirmed that cell proliferation and tumorigenicity of an HCC cell hepatocarcinogenesis can be caused by chronic hepatitis line in vitro. Sheng et al. [29] investigated how miR- B virus (HBV) infection [3]. Much effort towards the 101-3p regulated cell proliferation, cell cycle and apop- treatment of HBV-infected HCC has been made in the tosis in HCC and found that overexpression of miR- past, but with only limited success. Thus, identifying 101-3p caused an enhanced rate of apoptosis but no novel biochemical markers for early HCC diagnosis is a obvious change in the cell cycle. Besides, several onco- matter of the utmost urgency. genes, such as EZH2, FOS, COX-2 and SOX9, have miRNAs, ~ 20–22 nucleotides in length, are a class been found to be directly regulated by miR-101-3p/5p of small endogenous non-coding RNA molecules. [30–32]. Recently the potential of the miR-101 family They post-transcriptionally regulate mRNA expression as diagnostic indicators has also caught the eye of through imperfect base paring with the 30-untranslated researchers. He et al. [33] conducted a meta-analysis region of target genes. With comprehensive study, that summarized miRNAs’ diagnostic value in HCC miRNAs have become known as the star molecules of and found that miR-101-5p had great diagnostic value, cancer research. miRNAs in human cancers are though only three data sets were included and the involved in several pivotal biological processes (BP), results need to be further validated. Furthermore, in including cancer proliferation, differentiation, progres- human, miR-101 precursor transcripts are encoded sion and cell apoptosis [4–6]. Although their functions with two genomic loci (miR-101-1 and miR-101-2). remain elusive, up- and down-regulation of miRNAs For two mature miRNAs, miR-101-3p is generated have been widely reported in all kinds of cancer tissues from the 30 ends of the precursors, and miR-101-5p in comparison with expression in the corresponding from the 50 end of pre-miR-101-1 (http://www.mirbase. normal tissues [7,8]. In particular, miRNAs have been org/). We speculated that miR-101-3p may also serve found to be biomarkers for cancer clinical diagnosis, as a diagnostic marker for HCC. Since the seed region histological classification and prognosis [9–13]. of miR-101-3p and miR-101-5p is unique, they are pre- Accumulating evidence has clearly demonstrated dicted to regulate unique targets. However, to the best that the aberrant expression of miRNAs may further of our knowledge, the comparative roles of miR-101- influence the expression of tumor oncogenes and sup- 3p and miR-101-5p in HCC have not yet been fully pressor genes, thereby leading to the occurrence of a studied. tumor [14–17]. Theoretically, mature miRNA genera- The present study investigated miR-101-3p and tion requires a series of enzyme reactions. First, pri- miR-101-5p expression in HCC tissues compared with mary miRNA transcripts are cleaved in the nucleus by that in healthy controls. Published studies, Gene the Drosha enzyme to liberate the precursor miRNA Expression Omnibus (GEO) microarray chips and The (pre-miRNA) hairpin. Subsequently, the pre-miRNA Cancer Genome Atlas (TCGA) data that included is exported to the cytoplasm and further processed by miR-101-3p or miR-101-5p expression information the enzyme Dicer to produce two mature miRNAs were collected together. Additionally, previous studies (miR-5p and miR-3p) [18,19]. Even though the two have mainly focused on a single gene [34–36], and mature miRNAs are transcribed from the same pre- studies have rarely focused on the function of

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 65 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al. coexpressed genes in cancers. For the purpose of processed by the Dicer enzyme to form the mature miRNA. obtaining a full understanding of the molecular mecha- All of the available clinical parameters were analyzed by SPSS nisms underlying HCC, comprehensive bioinformatics STATISTICS 22.0 (IBM Corp., Armonk, NY, USA). methods were used to investigate the function and pathways of target genes of miR-101-3p and miR-101- Data mining 5p associated with HCC. In a word, the present study aimed to analyze the expression and mechanism of Search strategy and study selection miR-101-3p and miR-101-5p in the initiation and Comprehensive literature searches were conducted on elec- development of HCC. This exploration will provide tronic databases PubMed, EMBASE, Web of Science, the novel insights into HCC. A flowchart for the whole Cochrane Library, and Chinese National Knowledge study designed is shown in Fig. 1. Infrastructure (CNKI) up to 29 December 2016. No lan- guage limitations were imposed. Qualifying articles were Material and methods screened by combining the following keywords: ‘miR-101’ OR ‘miRNA-101’ OR ‘miRNA-101’ OR ‘miR101’ OR ‘miRNA101’ OR ‘miRNA 101’ OR ‘miR-101-5p’ OR The clinical role of miR-101 based on the public ‘miRNA-101-5p’ OR ‘miRNA-101-5p’ OR’miR-101-3p’ database TCGA OR ‘miRNA-101-3p’ OR ‘miRNA-101-3p’ AND malig- To verify the difference in the miR-101 expression levels nan* OR cancer OR tumor OR neoplas* OR carcinoma between HCC and normal liver tissues, we downloaded rele- AND hepatocellular OR liver OR hepatic OR HCC AND vant data from the public tumor database TCGA, in which diagnos* OR receiver operating characteristic (ROC) OR samples from 362 HCC patients and 50 adjacent non-HCC specificity OR sensitivity OR DEGs OR DEMs OR ‘differ- tissues were included. Additionally, miR-101-1 and miR-101- entially expressed’. In addition, the reference lists were also 2 levels were both calculated because the relevant sample data manually searched to reduce article omission. The title and were provided in TCGA. miR-101-1 and miR-101-2 are two abstract of the obtained studies were scanned to exclude precursor hairpin structures of miR-101 miRNA that are any clearly irrelevant publications. In addition to searching located in the on 1 (MI0000103) the literature, we also searched the GEO database for eligi- and 9 (MI0000739), respectively [37]. Both of them are ble microarrays with the following terms: malignan* OR

Fig. 1. Flowchart of the study design.

66 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

cancer OR tumor OR neoplas* OR carcinoma AND hepa- translated by MEDCALC 11.4.2.0 (MedCalc Software, Ostend, tocellular OR liver OR hepatic OR HCC. Belgium). To reduce inaccuracy in the relevant data extracted from the included studies, three independent researchers (XY, PL and JMC) performed the data extrac- Criteria for inclusion and exclusion tion separately. Studies that met the following criteria were included: (a) investigated HCC; (b) measured the level of miR-101, miR- Statistical analysis 101-3p or miR-101-5p in HCC tissue, plasma or serum; (c) included the diagnosis of HCC or the clinical parameters; All statistical analyses were performed using SPSS STATISTICS and (d) reported true positives (TPs), false positives (FPs), 20.0 or STATA 12.0 (StataCorp, College Station, TX, USA). false negatives (FNs), and true negatives (TNs) or sensitiv- For the clinical parameter analysis, miR-101 expression ity and specificity of miR-101. In addition, (e) if the studies was represented as the mean standard deviation. The did not provide a fourfold contingency table, they were standards for assessing the area under the curve (AUC) in included if the original data were available; and (f) the ROC curve were as follows: 0.5–0.7 represented poor microarrays were included if they enrolled more than three evidence for diagnosis, 0.7–0.9 represented moderate evi- patient samples and measured the miR-101 profile for dence for diagnosis and 0.9–1.0 represented high evidence HCC. for diagnosis. The correlation between miR-101 expression Articles that met the following criteria were excluded: and the clinicopathological parameters was investigated (a) studies without sufficient data, such as reviews or sys- with Spearman’s rank correlation. The significance of the tematic reviews, (b) repeat reports, (c) studies conducted on difference between HCC and non-cancerous liver tissues cell lines or animals and (d) letters to the editor or confer- was studied using Student’s t test. The significant differ- ence abstracts. ences among three groups were examined by one-way ANOVA. For data mining, the pooled sensitivity, speci- ficity, positive likelihood ratios (PLRs), negative likelihood Data synthesis and analysis ratios (NLRs), and diagnostic odds ratio with their corre- Studies that did not provide TPs, FPs, FNs and TNs but sponding 95% confidence intervals (CIs) were calculated gave sensitivity and specificity or the original data were with the bivariate regression model. Additionally, the

Table 1. Relationship between miR-101-1 and clinicopathological parameters in HCC (TCGA data).

miR-101-1 relative expression Correlation analysis

Clinicopathological feature n Mean SD tPrP

Tissue HCC 362 15.25 1.05 16.198 0.000 0.480 0.000 Normal 50 16.93 0.62 Gender Male 247 15.23 1.03 0.542 0.588 0.015 0.774 Female 115 15.29 1.09 Age < 50 years 68 15.06 1.13 1.708 0.089 0.031 0.559 ≥ 50 years 290 15.30 1.03 HBV 255 15.20 1.08 1.347 0.179 0.079 0.133 + 106 15.37 0.98 HCV 308 15.22 1.05 1.407 0.160 0.054 0.309 + 54 15.44 1.01 Pathological stage Stage I–II 250 15.34 1.01 3.276 0.001 0.174 0.001 Stage III–IV 88 14.92 1.16 Pathological T T1–T2 268 15.33 1.01 2.867 0.004 0.173 0.001 T3–T4 92 14.97 1.12 Histological grade GI–II 223 15.40 1.05 3.467 0.001 0.184 0.000 GIII–IV 135 15.00 1.02

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 67 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

Table 2. Relationship between miR-101-2 and clinicopathological parameters in HCC (TCGA data).

miR-101-2 relative expression Correlation analysis

Clinicopathological feature n Mean SD tPrP

Tissue HCC 362 15.27 1.05 16.256 0.000 0.481 0.000 Normal 50 16.94 0.61 Gender Male 247 15.25 1.03 0.533 0.594 0.016 0.760 Female 115 15.31 1.09 Age < 50 years 68 15.08 1.13 1.709 0.088 0.032 0.550 ≥ 50 years 290 15.32 1.03 HBV 255 15.22 1.07 1.347 0.179 0.079 0.133 + 106 15.39 0.98 HCV 308 15.24 1.05 1.425 0.155 0.054 0.309 + 54 15.46 1.01 Pathological stage Stage I–II 250 15.36 1.01 3.309 0.001 0.174 0.001 Stage III–IV 88 14.93 1.15 Pathological T T1–T2 268 15.35 1.00 2.904 0.004 0.174 0.001 T3–T4 92 14.99 1.12 Histological grade GI–II 223 15.41 1.04 3.452 0.001 0.184 0.000 GIII–IV 135 15.03 1.02

inconsistency were used to quantify heterogeneity between studies [39]. The possibility of publication bias was finally explored by Deeks’ funnel plot, and P values < 0.1 were considered significant.

Bioinformatic analysis In silico analysis of target genes of miR-101

MiRWalk2.0 [40] (http://zmf.umm.uni-heidelberg.de/apps/ zmf/mirwalk2/), which combines 12 existing miRNA-target prediction programs, was used to provide comprehensive potential targets for miR-101-3p and miR-101-5p. The genes identified by more than eight prediction software pro- grams for miR-101-3p and more than six for miR-101-5p were selected to obtain more reliable targets.

Natural language processing

Natural language processing (NLP) is a novel computerized Fig. 2. MiR-101 expression profiles for the diagnosis of HCC. The approach to analyze electronic free text to achieve ‘human- AUC of the low miR-101 level for HCC diagnosis was 0.925 (95% like language processing’. With this approach, program- CI: 0.896–0.953, P < 0.001). mers create software to ‘read’ text and extract key pieces of information from clinician notes, procedure/radiology/ summary receiver operator characteristic (SROC) curve pathology reports and laboratory results [41,42]. We per- with the area under the SROC curve was calculated [38]. formed a literature search in PubMed to obtain all related What is more, the Q test and the I2 measure of electronic records. The detailed process was described in

68 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC our previous article [43,44]. Finally, 1800 genes that were analysis. The GO function analysis categorized selected differentially expressed in HCC were identified for further genes into groups in accordance with three independent analysis. classification standards, BPs, cellular components (CCs), and molecular functions (MFs). The top 10 terms of each GO category and top 30 pathways of the KEGG pathways Functional and signaling pathway analyses were visualized as GO maps and KEGG maps, separately, A set of condition-specific genes from the overlapping via CYTOSCAPE v3.4.0 (http://cytoscape.org/). genes from the target prediction software and NLP further underwent functional and signaling pathway analyses on a Protein–protein interaction network construction public database platform, the Database for Annotation, Visualization and Integrated Discovery (DAVID; https://da Overlapping genes were inputted to the STRING v10.0 online vid.ncifcrf.gov/), which provides a functional interpretation tool (http://string-db.org/) to construct the protein–protein of massive gene lists derived from genomic studies. The interaction (PPI) network. The direct (physical) and indi- analyses included Gene Ontology (GO) function analysis rect (functional) associations of were derived from (http://www.geneontology.org/) and Kyoto Encyclopedia of four methods: (a) literature-reported protein interactions, Genes and Genomes (KEGG; http://www.genome.jp/kegg/) (b) high-throughput experiments, (c) genome analysis and

Fig. 3. Diagnostic accuracy of miR-101-3p in HCC. (A) Sensitivity (SENS) and specificity (SPEC) with corresponding heterogeneity statistics. (B) SROC curves for miR-101-3p with CI in the diagnosis of HCC.

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 69 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al. prediction and (d) coexpression studies. By scrutinizing the for miR-101-2 (15.27 1.05 in HCC and 16.94 0.61 connectivity degrees of the nodes in the PPI networks, we in para-non-cancerous liver tissues, P < 0.001). Addi- determined the hub genes. A node with a high degree of tionally, the altered expressions of miR-101-1 and connectivity is perceived as a hub node. miRNA-101-2 were both associated with pathological stage, pathological T stage and histological stage. Results Compared with the expression in advanced stage (III and IV) HCC patients, the relative expression of miR- Clinicopathological significance of miR-101-1/ 101 in early stage patients was notably increased (I < miR-101-2 in HCC tissues and II, P 0.05), and the Spearman correlation test confirmed that the correlations between miR-101 and The relationship between miR-101-1/miR-101-2 and the pathological stage, pathological T stage and histo- clinicopathological parameters in HCC was mined logical stage were r = 0.17, P = 0.001; r = 0.17, from TCGA, as shown in Tables 1 and 2. Data profil- P = 0.001 and r = 0.18, P < 0.001, respectively. ing revealed that when compared with the expression in para-non-cancerous normal tissues (16.93 0.62), miR-101-1 expression was significantly reduced in Diagnostic value of miR-101-3p and miR-101-5p HCC tissues (15.25 1.05, P < 0.001). In addition, Study selection the AUC of the low miR-101-1 level for HCC diagno- sis was 0.925 (95% CI: 0.896–0.953, P < 0.001, Fig. 2) Through the literature search, 341 relevant articles with a cut-off value of 16.17 (82.6% sensitivity and were identified, 339 of which were excluded for being 90.0% specificity). Similar results were also obtained case reports, reviews, letters, repeat publications and

Fig. 4. Fagan diagram and likelihood matrix for miR-101-3p to diagnose cancer or to eliminate the diagnosis of cancer. (A) Pre-test probability of the miR-101-3p assay in HCC detection. (B) Likelihood matrix showing individual (circles) and pooled (diamond) values of PLRs combined with NLRs. LLQ, left lower quadrant; LUQ, left upper quadrant; RLQ, right lower quadrant; RUQ, right upper quadrant.

70 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

Fig. 5. Diagnostic accuracy of miR-101-5p in HCC. (A) Sensitivity (SENS) and specificity (SPEC) with corresponding heterogeneity statistics. (B) SROC curves for miR-101-5p with CI in the diagnosis of HCC. studies not specifically pertaining to miR-101-3p/5p. HCC and 114 normal control samples were down- The two remaining publications were examined by loaded from the GEO database to calculate the miR- three researchers and ultimately included. Moreover, 101-5p diagnostic value (GSE74618, GSE21362, GEO microarrays that detected miR-101-3p and/or GSE40744, GSE41874 and GSE57555). In addition, miR-101-5p were identified for further analysis and the precursors of miR-101 identified from TCGA were were combined after assessment. Finally, 12 datasets also considered. including 315 HCC and 330 normal control samples were downloaded from the GEO database to calculate Heterogeneity analysis the miR-101-3p diagnostic value (GSE39678, GSE21279, GSE67882, GSE65708, GSE12717, GSE The analysis of heterogeneity is widely used to 10694, GSE22058, GSE21362, GSE40744, GSE41874, evaluate the accuracy of statistical pooling from GSE54751 and GSE57555); five datasets including 308 multiple studies [45]. Since heterogeneities may come

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 71 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

Fig. 6. Fagan diagram and likelihood matrix for miR-101-5p to diagnose cancer or to eliminate the diagnosis of cancer. (A) Pre-test probability of the miR-101-5p assay in HCC detection. (B) Likelihood matrix showing individual (circles) and pooled (diamond) values of PLRs combined with NLRs. LLQ, left lower quadrant; LUQ, left upper quadrant; RLQ, right lower quadrant; RUQ, right upper quadrant.

Fig. 7. The Deeks’ test that assesses potential publication bias in the miR-101 assay. (A) Potential publication bias assessment of miR-101- 3p. (B) Potential publication bias assessment of miR-101-5p. from a threshold effect and a non-threshold effect, the among the included studies. In other words, the corre- threshold effect was first explored by the Spearman lation coefficient and P value between the logit of sen- test to calculate the heterogeneity of miR-101-3p/5p sitivity and logit of 1 – specificity were calculated. As

72 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

Fig. 8. The KEGG pathway analysis of miR-101-3p predicted target genes in HCC. Pathway analyses were performed to identify significantly enriched pathways by using CYTOSCAPE v3.4.0. The top 30 pathways are displayed; the map node size represents the P value of targets, low values are indicated by large nodes, and the node color represents the gene count number with low values indicated by pink.

Fig. 9. The KEGG pathway analysis of miR-101-5p predicted target genes in HCC. Pathway analyses were performed to identify significantly enriched pathways by using CYTOSCAPE v. 3.4.0. The top 30 pathways are displayed; the map node size represents the P value of targets, low values are indicated by large nodes, and the node color represents the gene count number with low values indicated by pink. a result, the Spearman correlation coefficients for Diagnostic accuracy of miR-101-3p and miR-101-5p in miR-101-3p and miR-101-5p were 0.386 (P = 0.215) HCC and 0.059 (P = 0.912), respectively, indicating that heterogeneity from the threshold effect was not found. Evident heterogeneity for pooled sensitivity and speci- However, the I2 values in the forest plots of sensitivity ficity of miR-101-3p was seen in the collected data and specificity (more than 50%) revealed that we can- (I2 = 87.02% and 83.73%, respectively, P < 0.05); not ignore the non-threshold effect from the included thus, a random effects model was finally selected, studies. based on which the pooled sensitivity and specificity of

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 73 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

Table 3. KEGG functional annotation for most significantly related targets of miR-101.

KEGG ID Term Gene no. P Genes

MiR-101-3p hsa04520 Adherens junction 5 8.53 9 104 MAP3K7, MAPK1, TGFBR1, NLK, SSX2IP hsa05140 Leishmaniasis 5 8.53 9 104 MAP3K7, MAPK1, FOS, PTGS2, JAK2 hsa04917 Prolactin signaling pathway 5 8.53 9 104 MAPK1, FOS, SOCS2, GSK3B, JAK2 hsa05200 Pathways in cancer 9 2.18 9 103 CEBPA, MAPK1, FOS, PTGS2, TGFBR1, GSK3B, PTCH1, CDK6, CXCL12 hsa04010 MAPK signaling pathway 7 4.09 9 103 MAP3K7, MAPK1, FOS, DUSP1, TGFBR1, NLK, SRF hsa05210 4 6.16 9 103 MAPK1, FOS, TGFBR1, GSK3B hsa04360 Axon guidance 5 7.10 9 103 MAPK1, EPHA7, NRP1, GSK3B, CXCL12 hsa05162 Measles 5 8.34 9 103 MAP3K7, GSK3B, IL13, CDK6, JAK2 hsa04068 FoxO signaling pathway 5 8.56 9 103 MAPK1, TGFBR1, NLK, CCNG2, BCL2L11 hsa05166 HTLV-I infection 6 1.86 9 102 FOS, NRP1, ETS1, TGFBR1, GSK3B, SRF MiR-101-5p hsa05200 Pathways in cancer 35 7.08 9 1016 XIAP, PTGS2, FOXO1, MMP1, TPM3, CCNE2, IGF1R, KRAS, CDKN2B, BCL2, SOS1, ITGAV, , AKT3, PRKCA, BMP4, IL6, RALBP1, TGFBR1, CREBBP, et al. hsa04068 FoxO signaling pathway 21 5.59 9 1014 IL6, SGK3, TGFBR1, CREBBP, SMAD4, SMAD3, FOXO1, SMAD2, MAPK10, IL7R, CCNG2, BCL2L11, ATM, IGF1R, NRAS, MAPK1, KRAS, CDKN2B, CCND2, SOS1, AKT3 hsa05161 Hepatitis B 20 2.82 9 1012 PRKCA, IL6, YWHAZ, TGFBR1, CREBBP, MAP2K4, CYCS, SMAD4, CDK6, MAPK10, STAT2, CCNE2, NRAS, MAPK1, KRAS, DDX3X, BCL2, NFATC2, MYC, AKT3 hsa04151 PI3K–Akt signaling pathway 27 7.91 9 1011 YWHAZ, RPS6KB1, IL7R, CCNE2, IGF1R, KRAS, SOS1, ITGAV, BCL2, ANGPT2, MYC, AKT3, GHR, PRKCA, IL6, FLT1, SGK3, MET, CDK6, BCL2L11, et al. hsa04917 Prolactin signaling pathway 12 2.43 9 108 MAPK1, NRAS, KRAS, TNFRSF11A, SOCS2, PRLR, CCND2, SOS1, ESR1, ESR2, MAPK10, AKT3 hsa04520 Adherens junction 12 2.43 9 108 MAP3K7, MAPK1, IGF1R, TGFBR1, CREBBP, MET, SMAD4, SMAD3, SMAD2, WASL, YES1, CTNNA1 hsa05220 Chronic myeloid leukemia 12 2.83 9 108 MAPK1, NRAS, KRAS, CRKL, HDAC2, TGFBR1, SOS1, SMAD4, CDK6, MYC, AKT3, PTPN11 hsa05205 Proteoglycans in cancer 18 3.80 9 108 PRKCA, ERBB4, MET, ESR1, RPS6KB1, SDC2, FZD7, PTPN11, IGF1R, NRAS, MAPK1, KRAS, ITGAV, SOS1, MYC, FRS2, AKT3, TWIST1 hsa05210 Colorectal cancer 11 7.32 9 108 MAPK1, KRAS, TGFBR1, BCL2, CYCS, SMAD4, SMAD3, SMAD2, MAPK10, MYC, AKT3 hsa04630 Jak–STAT signaling pathway 15 1.37 9 107 IL6, SOCS2, IL6ST, LEPR, CREBBP, IL7R, STAT2, PTPN11, LEP, PRLR, CCND2, SOS1, MYC, AKT3, GHR miR-101-3p was 78.0% (95% CI: 65.0–88.0%) and With regard to miR-101-5p, a random effects model 79.0% (95% CI: 67.0–88.0%), respectively (Fig. 3A). was also selected for further analysis (I2 = 78.79% and In addition, the summary SROC of miR-101-3p was 92.24%, for sensitivity and specificity, respectively, 0.86 (95% CI: 0.82–0.89; Fig. 3B). In addition, the P < 0.01). The overall sensitivity and specificity were PLR and NLR for HCC diagnosis were 3.803 (95% 79.0% (95% CI: 75.0–83.0%) and 60.0% (95% CI: CI: 2.398–6.033) and 0.272 (95% CI: 0.169–0.439), 27.0–86.0%), respectively (Fig. 5A). The calculated respectively. Furthermore, the pre-test probability was AUC of the summary SROC was 0.80 (95% CI: 0.76– 20 for miR-101-3p, and the corresponding positive and 0.83; Fig. 5B). Additionally, the PLR was 1.981 (95% negative post-test probabilities of miR-101-3p were 49 CI: 0.831–4.721), and the NLR was 0.349 (95% CI: and 6, respectively, suggesting that the power of miR- 0.174–0.698); the pre-test probability of miR-101-5p 101-3p to diagnose real patients as HCC was 3.8 times was 20, and the corresponding positive and negative the normal control (Fig. 4A). In addition, the likeli- post-test probability of miR-101-5p was 33 and 8, hood ratio scattergram disclosed that the summary respectively, suggesting that the power of miR-101-5p to point of the PLR together with the NLR lies in the diagnose real patients as HCC was 1.98 times the nor- right lower quadrant (PLR < 10, NLR > 0.1; Fig. 4B). mal control (Fig. 6A). In addition, the summary point

74 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

Table 4. GO functional annotation for most significantly related targets of miR-101-3p.

Gene GO ID Term no. P Genes

BP GO: 0045944 Positive regulation of 20 1.90 9 108 CEBPA, SOX6, ZEB1, ZIC1, TET2, SOX9, PROX1, SRF, from RNA polymerase II promoter MYCN, PGR, MEF2D, FOS, ETS1, ZNF148, GSK3B, ASH1L, NEUROD1, TCF4, BCL9, SMARCA4 GO: 0045893 Positive regulation of transcription, 15 3.19 9 108 KLF6, RSF1, TGFBR1, ARID1A, SOX9, ZIC1, PROX1, DNA-templated MYCN, MAPK1, FOS, ETS1, NEUROD1, PTCH1, TCF4, SMARCA4 GO: 0032355 Response to estradiol 6 4.40 9 105 DUSP1, SOCS2, PTGS2, ETS1, EZH2, PTCH1 GO: 0008285 Negative regulation of cell 10 4.72 9 105 CEBPA, PTGS2, ETS1, CDK6, JAK2, ZEB1, ARID2, PROX1, proliferation SRF, CDH5 GO: 0001764 Neuron migration 6 8.71 9 105 NRP1, GJA1, PAFAH1B1, TOP2B, CXCL12, SRF GO: 0001701 In utero embryonic development 7 1.49 9 104 TGFBR1, MYO1E, GJA1, PTCH1, SOX6, SRF, BCL2L11 GO: 0000122 Negative regulation of transcription 12 2.32 9 104 CUL3, CEBPA, JDP2, ZNF148, EZH2, PTCH1, ARID1A, from RNA polymerase II promoter SOX6, ZEB1, TCF4, PROX1, SMARCA4 GO: 0006366 Transcription from RNA polymerase 10 3.32 9 104 CEBPA, MEF2D, FOS, ZNF148, ETS1, ASH1L, NEUROD1, II promoter ZIC1, SOX9, SRF GO: 0018107 Peptidyl-threonine phosphorylation 4 5.73 9 104 MAPK1, TGFBR1, GSK3B, NLK GO: 0007179 Transforming growth factor beta 5 6.53 9 104 MAP3K7, FOS, TGFBR1, NLK, CDH5 signaling pathway CC GO: 0005654 Nucleoplasm 30 2.50 9 107 ING3, RSF1, XPO5, XPO4, EZH2, ZEB1, SOX6, ZIC1, SOX9, SRF, ARID2, LARP1, CUL3, PGR, FOS, FBXW7, ZNF148, TOP2B, BCL9, NLK, et al. GO: 0005634 Nucleus 40 1.24 9 105 ING3, JDP2, RSF1, PTGS2, XPO5, EZH2, SOX6, ZEB1, SOX9, ZIC1, TIMP3, SRF, MAP3K7, CUL3, PGR, FOS, MSI1, SSX2IP, TOP2B, TCF4, et al. GO: 0009897 External side of plasma 6 1.54 9 103 CLCN3, FGA, IL13, ABCA1, CXCL12, CDH5 membrane GO: 0005901 Caveola 4 2.17 9 103 MAPK1, PTGS2, PTCH1, JAK2 GO: 0009986 Cell surface 8 5.50 9 103 CLCN3, NRP1, FGA, TGFBR1, CFTR, SPARC, CDH5, SLC7A11 GO: 0043234 Protein complex 7 5.73 9 103 MAPK1, FBXW7, PTGS2, CFTR, SSX2IP, SOX9, SMARCA4 GO: 0000790 Nuclear 5 7.16 9 103 EZH2, ARID1A, TCF4, SRF, SMARCA4 GO: 0005737 Cytoplasm 31 1.21 9 102 ING3, PTGS2, XPO5, XPO4, EZH2, IL13, ZEB1, ZIC1, CCNG2, SRF, LIN28B, LARP1, MAP3K7, FBXW7, MSI1, TOP2B, ZMYM2, SOCS2, MYO1E, CFTR, et al. GO: 0005769 Early endosome 5 1.28 9 102 MAPK1, CLCN3, NRP1, GJA1, CFTR GO: 0005794 Golgi apparatus 9 2.02 9 102 CUL3, MAPK1, CLCN3, ZNF148, ASH1L, GJA1, PTCH1, ABCA1, BCL9 MF GO: 0005515 Protein binding 62 1.67 9 109 JDP2, NRP1, RSF1, PTGS2, XPO5, XPO4, EZH2, IL13, GJA1, ZEB1, LARP1, MAP3K7, PGR, CUL3, FOS, ZNF148, SOCS2, CFTR, ARID1A, CDK6, et al. GO: 0000978 RNA polymerase II core promoter 11 2.20 9 106 PGR, CEBPA, MEF2D, FOS, JDP2, ZNF148, NEUROD1, proximal region sequence-specific TCF4, ZIC1, SRF, SMARCA4 DNA binding GO: 0001077 Transcriptional activator activity, 9 6.46 9 106 PGR, CEBPA, MEF2D, FOS, NEUROD1, TCF4, ZIC1, SOX9, RNA polymerase II core promoter SRF proximal region sequence-specific binding GO: 0008134 binding 9 2.47 9 105 CEBPA, MAPK1, FOS, ETS1, NLK, NEUROD1, ZEB1, SRF, SMARCA4

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 75 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

Table 4. (Continued).

Gene GO ID Term no. P Genes

GO: 0003700 Transcription factor activity, 15 3.92 9 105 CEBPA, JDP2, SOX6, ZEB1, SOX9, ZIC1, PROX1, SRF, sequence-specific DNA binding MYCN, PGR, MEF2D, FOS, ETS1, NEUROD1, TCF4 GO: 0003677 DNA binding 20 4.29 9 105 CEBPA, KLF6, ZMYM2, EZH2, ARID1A, SOX6, ZEB1, TET2, ARID2, LIN28B, PROX1, MYCN, PGR, MAPK1, FOS, SP2, ETS1, ASH1L, TOP2B, TCF4 GO: 0030332 Cyclin binding 4 7.72 9 105 CUL3, FBXW7, PTCH1, CDK6 GO: 0046982 Protein heterodimerization activity 10 1.38 9 104 CUL3, MEF2D, FOS, CLCN3, JDP2, NEUROD1, SOX6, TOP2B, TCF4, SOX9 GO: 0003682 Chromatin binding 9 2.29 9 104 FOS, JDP2, EZH2, ASH1L, NEUROD1, ZEB1, TOP2B, TCF4, SOX9 GO: 0005102 Receptor binding 8 6.95 9 104 PGR, FGA, CADM2, GJA1, JAK2, ABCA1, CXCL12, CDH5

of the PLR combined with the NLR also lies in the right pathways in cancer (hsa05200: P = 0.002) and (e) the lower quadrant (PLR < 10, NLR > 0.1), which was mitogen-activated protein kinase (MAPK) signaling consistent with the Fagan’s nomogram result (Fig. 6B). pathway (hsa04010: P = 0.004). For miR-101-5p, 90 KEGG pathways were enriched, and the top pathways were (a) pathways in cancer (hsa05200: Publication bias P = 7.08 9 10 16), (b) the forkhead box O (FoxO) Publication bias was conducted by using the Deeks’ signaling pathway (hsa04068: P = 5.59 9 1014), funnel plot asymmetry test. According to the results, (c) hepatitis B (hsa05161: P = 2.82 9 1012), (d) the the funnel plots that represented every study were phosphoinositide-3-kinase (PI3K)–Akt signaling almost symmetric, suggesting that publication bias pathway (hsa04151: P = 7.91 9 1011) and (e) the from the studies included was absent in our study. The prolactin signaling pathway (hsa04917: P = 2.43 9 obtained P-values of 0.718 and 0.447 for miR-101-3p 108). The top 30 pathways associated with miR-101- and miR-101-5p, respectively, also revealed the 3p and miR-101-5p are shown in Figs 8 and 9, respec- absence of publication bias (Fig. 7). tively, and all of the pathways are displayed in Table 3. Bioinformatic analysis GO enrichment analysis To improve understanding of the function of miR-101, the potential target genes of miR-101-3p and miR-101- As shown in Tables 4 and 5, the GO enrichment was 5p in HCC were identified separately. Based on the composed of the BP, CC and MF categories. In the prediction software and NLP, 73 target genes corre- BP category for miR-101-3p, we can observe that the sponding to miR-101-3p and 90 target genes corre- 73 target genes were mainly enriched in (a) positive sponding to miR-101-5p were obtained. Subsequently, regulation of transcription from the RNA poly- bioinformatic analyses were conducted to investigate meraseIIpromoter (GO: 0045944, P = 1.90 9 108), the function and pathways of target genes of miR-101 (b) positive regulation of transcription, DNA-tem- associated with HCC. All of the target genes were plated (GO: 0045893, P = 3.19 9 108) and (c) inputted into DAVID for bioinformatic analysis. response to estradiol (GO: 0032355, P = 4.40 9 105). In addition, in the CC category, (a) nucleoplasm (GO: 0005654, P = 2.50 9 107), (b) nucleus (GO: 0005634, KEGG pathway enrichment analysis P = 1.24 9 10 5) and (c) external side of plasma mem- Our study revealed that 23 KEGG pathways corre- brane (GO: 0009897, P = 0.002) remained the top sponding to miR-101-3p were enriched, from which three enriched items. With regard to MF, the top the top five pathways in which target genes were ranking three items were protein binding (GO: enriched were (a) the adherens junction pathway 0005515, P = 1.67 9 109), RNA polymerase II core (hsa04520: P = 8.53 9 104), (b) the leishmaniasis promoter proximal region sequence-specific DNA pathway (hsa05140: P = 8.53 9 104), (c) the prolactin binding (GO: 0000978, P = 2.20 9 106) and tran- signaling pathway (hsa04917: P = 8.53 9 104), (d) scriptional activator activity, RNA polymerase II core

76 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

Table 5. GO functional annotation for most significantly related targets of miR-101-5p.

Gene GO ID Term no. P Genes

BP GO: 0043066 Negative regulation of apoptotic 29 2.01 9 1013 YWHAZ, MTDH, ERBB4, XIAP, IL6ST, FOXO1, PRKDC, process RPS6KB1, IGF1R, DDX3X, BCL2, TPT1, GLO1, MYC, TWIST1, BMP4, IL6, TBX3, SOCS2, SMAD3, et al. GO: 0045893 Positive regulation of transcription, 28 2.34 9 1011 RSF1, ERBB4, FOXO1, ZIC1, ASPH, NFATC2, MYC, BMP4, DNA-templated KLF6, IL6, TBX3, TGFBR1, CREBBP, SMAD5, SMAD4, ESR1, ATAD2, SMAD3, SMAD2, ESR2, et al. GO: 0045944 Positive regulation of transcription 38 5.43 9 1011 PRKDC, FOXO1, ZEB2, NR3C1, SOX6, ZEB1, ZIC1, PGR, from RNA polymerase II promoter IL17A, BARX2, CDKN2B, DDX3X, ZNF148, NFATC2, YES1, MYC, TWIST1, CKAP2, BMP4, IL6,et al. GO: 0000122 Negative regulation of transcription 32 9.99 9 1011 JDP2, MTDH, USP2, FOXO1, ZEB2, SOX6, ZEB1, BARX2, from RNA polymerase II promoter ZNF148, NFATC2, MYC, TWIST1, BMP4, DAB2IP, TBX3, YY1, CREBBP, SMAD4, ESR1, KLF17, et al. GO: 0008284 Positive regulation of cell 25 4.77 9 1010 ERBB4, IL6ST, IGF1R, CD47, KRAS, TNFRSF11A, ITGAV, proliferation BCL2, MYC, IL6, FLT1, KLB, TBX3, TGFBR1, PROX1, TET1, LEP, MAPK1, HDAC2, CRKL, et al. GO: 0043065 Positive regulation of apoptotic 19 7.80 9 109 BMP4, IL6, DAB2IP, ERBB4, PTGS2, PRKDC, FOXO1, process FRZB, LATS1, BCL2L11, ATM, BAK1, TRIM35, ITGA6, DDX3X, SFRP1, ATG7, SOS1, UNC5C GO: 0050900 Leukocyte migration 12 1.23 9 107 NRAS, CD47, KRAS, ITGA6, ITGAV, SOS1, TREM1, YES1, ANGPT2, MMP1, SLC7A11, PTPN11 GO: 0008285 Negative regulation of cell 18 2.36 9 106 BMP4, IL6, DAB2IP, ERBB4, PTGS2, SMAD4, SMAD2, proliferation CDK6, ZEB1, FRZB, ARID2, PROX1, SLIT3, BAK1, SPRY1, CDKN2B, SFRP1, MDM4 GO: 0001568 Blood vessel development 7 3.73 9 106 MIB1, LAMA4, CRKL, TBX3, ITGAV, FOXO1, AHR GO: 0071498 Cellular response to fluid shear 5 7.02 9 106 MTSS1, PTGS2, CA2, NFE2L2, TFPI2 stress CC GO: 0005829 Cytosol 67 3.59 9 108 RPL36A, FOXO1, RPS6KB1, LATS1, MAP3K7, CCNE2, BAK1, SPRY1, GSTM3, CDKN2B, MAT1A, ATG7, MYC, PRKCA, DAB2IP, SOCS2, SGK3, RALBP1, G3BP1, CYCS, et al. GO: 0005654 Nucleoplasm 53 1.23 9 105 RSF1, XPO4, FOXO1, RPS6KB1, ZEB1, ZIC1, PGR, CCNE2, SPRY1, CDKN2B, ZNF148, MYC, AKT3, PRKCA, DTL, ESR1, CDK6, ESR2, AHR, MCM6, et al. GO: 0009897 External side of plasma membrane 12 1.50 9 105 EPHA5, VCAM1, CLCN3, IL17A, IL6, TNFRSF11A, ITGA6, CD40LG, IL6ST, ITGAV, CD274, IL7R GO: 0005634 Nucleus 84 2.85 9 105 JDP2, RSF1, PTGS2, CPEB4, FOXO1, ZEB2, RPS6KB1, ZEB1, ZIC1, MAP3K7, CCNE2, PGR, GSTM3, BARX2, CDKN2B, TPT1, LOX, TFPI2, MYC, ANGPT2, et al. GO: 0005737 Cytoplasm 81 4.65 9 105 MTSS1, PTGS2, XPO4, CPEB4, FOXO1, RPS6KB1, ZEB1, ZIC1, MAP3K7, SPRY1, GSTM3, CDKN2B, ATG7, TPT1, FRS2, AKT3, PRKCA, DAB2IP, SOCS2, LPGAT1, et al. GO: 0071141 SMAD protein complex 4 6.01 9 105 SMAD5, SMAD4, SMAD3, SMAD2 GO: 0005667 Transcription factor complex 10 1.97 9 104 BARX2, YY1, SMAD5, SMAD4, SMAD3, PRKDC, SMAD2, DACH1, ZEB1, AHR GO: 0043235 Receptor complex 8 3.74 9 104 IGF1R, FLT1, ERBB4, LEPR, TGFBR1, NTRK2, SMAD3, GHR GO: 0009986 Cell surface 16 5.76 9 104 CLCN3, TGFBR1, MET, RPS6KB1, CFTR, SDC2, SLC7A11, VCAM1, ITGA6, SULF2, PRLR, SFRP1, CD40LG, ITGAV, CNTN2, GHR GO: 0005622 Intracellular 27 1.45 9 103 PRKCA, KLB, TGFBR1, G3BP1, SMAD5, SOCS6, SMAD4, DCDC2, SMAD3, MAPK10, LATS1, SEC63, WSB1, MAPK1, RND3, NRAS, TRIM35, KRAS, SFRP1, TRIM33, et al.

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 77 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

Table 5. (Continued).

Gene GO ID Term no. P Genes

MF GO: 0005515 Protein binding 147 5.59 9 1014 MTSS1, RPL36A, JDP2, PTGS2, IL6ST, XPO4, FOXO1, RPS6KB1, PGR, MAP3K7, CD47, BAK1, CDKN2B, ATG7, TPT1, ASPH, LOX, FRS2, TWIST1, DAB2IP, et al. GO: 0008134 Transcription factor binding 18 1.69 9 108 YWHAZ, CREBBP, ESR1, SMAD3, PRKDC, SMAD2, ZEB1, AHR, MAPK1, HDAC2, SP1, DDX3X, PSMD10, ATG7, BCL2, NFATC2, MYC, TWIST1 GO: 0001078 Transcriptional repressor activity, 10 3.79 9 106 JDP2, TBX3, ZNF148, YY1, CREBBP, KLF17, FOXO1, RNA polymerase II core promoter DACH1, NFATC2, PROX1 proximal region sequence-specific binding GO: 0005524 ATP binding 37 5.87 9 106 CLCN3, ERBB4, PFKFB2, STK17B, PRKDC, RPS6KB1, LATS1, MAP3K7, IGF1R, KRAS, DDX3X, MAT1A, HSPE1, YES1, AKT3, PRKCA, FLT1, SGK3, TGFBR1, UBE4B, et al. GO: 0000978 RNA polymerase II core promoter 15 4.05 9 105 JDP2, TBX3, SMAD4, ESR1, KLF17, SMAD3, SMAD2, proximal region sequence-specific NR3C1, ZIC1, PGR, HDAC2, SP1, ZNF148, NFATC2, MYC DNA binding GO: 0042802 Identical protein binding 22 8.15 9 105 MTSS1, YWHAZ, DAB2IP, XIAP, USP2, SMAD4, ESR1, SMAD3, CLDN10, RPS6KB1, STAT2, MCM6, PBLD, IGF1R, BAK1, MAPK1, GLUL, GSTM3, SFRP1, BCL2, CNTN2, GBP1 GO: 0008270 Zinc ion binding 29 8.19 9 105 RSF1, XIAP, NR3C1, ZEB1, LIN28B, MMP1, PGR, GLO1, XAF1, RCHY1, PRKCA, ZMYM2, YY1, CREBBP, ESR1, SMAD3, WHSC1, ESR2, TET2, TET1, et al. GO: 0004672 Protein kinase activity 14 1.79 9 104 PRKCA, SGK3, TGFBR1, MET, MAP2K4, STK17B, PRKDC, RPS6KB1, PBK, MAPK10, MAP3K7, MAP4K4, HIPK2, AKT3 GO: 0043565 Sequence-specific DNA binding 17 2.04 9 104 JDP2, TBX3, SMAD4, ESR1, SMAD3, FOXO1, WHSC1, NR3C1, ESR2, SOX6, PGR, HDAC2, SP1, ZNF148, BCL2, NFE2L2, MYC GO: 0019899 Enzyme binding 13 3.32 9 104 PRKCA, PTGS2, UBE4B, ESR1, PRKDC, CFTR, ESR2, PGR, GSTM3, HDAC2, MDM4, YES1, GBP1

promoter proximal region sequence-specific binding binding (GO: 0008131, P = 1.69 9 108) and (c) tran- (GO: 0001077, P = 6.46 9 106). In the same way, the scriptional repressor activity, RNA polymerase II core top three enriched pathways of the 119 target genes of promoter proximal region sequence-specific binding miR-101-5p in the BP category were (a) negative regu- (GO: 0001078. P = 3.79 9 106). All of the GO lation of the apoptotic process (GO: 0043066, enrichment items are visualized in the GO network P = 2.01 9 1013), (b) positive regulation of transcrip- (Figs 10 and 11). tion, DNA-templated, negative regulation of transcrip- tion from the RNA polymerase II promoter (GO: Protein–protein interaction network 0045893, P = 2.34 9 10 11) and (c) positive regulation of transcription from the RNA polymerase II pro- A PPI network was designed to screen out the hub moter (GO: 0045944, P = 5.43 9 1011). For the CC genes according to the degree to which each of the category, (a) cytosol (GO: 0005829, P = 3.59 9 108), genes appeared in the network. Here, the PPI network (b) nucleoplasm (GO: 0005654, P = 1.23 9 105) and was constructed by using the STRING database. As (c) external side of the plasma membrane (GO: shown in Figs 12 and 13, FOX, SMARCA4 and 0009897, P = 1.50 9 105) were the most enriched MAPK1 remained the top three utmost important terms. In addition, in the MF category, the utmost sig- genes for miR-101-3p, while ESR1, KRAS, NRAS, nificant three items were (a) protein binding (GO: FOXO1, CREBBP and SMAD3 were regarded as the 0005515, P = 5.59 9 1014), (b) transcription factor hub genes for miR-101-5p.

78 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

Fig. 10. GO functional analysis of miR- 101-3p in HCC. Top 10 terms of each category are displayed, and every node represents different BP terms; the map node size represents the P value of targets, low values are indicated by large nodes, and the node color represents the gene count number with low values indicated by pink.

Fig. 11. GO functional analysis of miR-101-5p in HCC. Top 10 terms of each category are displayed, and every node represents different BP terms; the map node size represents the P value of targets, low values are indicated by large nodes, and the node color represents the gene count number with low values indicated by pink.

Discussion parameters. TCGA data showed that the miR-101 level was significantly lower in HCC than in In the present study, we investigated the relationship para-non-cancerous liver tissues, and great diagnostic between miR-101 expression and clinicopathological value of miR-101 in HCC was found. Additionally,

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 79 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

limited number of available publications, the exact diagnostic value of miR-101 and the difference between miR-101-3p and miR-101-5p are still unclear. Alpha-fetoprotein (AFP), as the traditional marker of liver diseases, has been used for HCC diagnosis in the clinic. Recently, He et al. [33] conducted a meta-analy- sis with 10 data sets (879 HCC patients and 1028 con- trols) assessing AFP for HCC diagnosis and revealed that the AUC–SROC of pooled AFP was 0.82 (95% CI: 0.78–0.85), with sensitivity of 0.631 (95% CI: 0.552–0.703) and specificity of 0.943 (95% CI: 0.875– 0.975). Here, we first combined gene expression microarray datasets from the GEO database and RNA-seq from TCGA database, as well as two stud- ies, to further confirm the diagnostic efficacy of miR- 101-3p and miR-101-5p and then discover the differ- ence between the two mature mRNAs. Our findings Fig. 12. The PPI network of miR-101-3p potential targets. Both the suggested that the pooled diagnostic accuracy of miR- color and the size of the nodes reflect the connectivity degrees of 101-3p for HCC (SROC: 0.86 (95% CI: 0.82–0.89); two nodes; nodes with a green color are perceived as hub genes. sensitivity and specificity were 78.0% (95% CI: 65.0– miR-101 was negatively correlated with pathological 88.0%) and 79.0% (95% CI: 0.67–88.0%), respec- stage, pathological T stage and histological grade. tively), which showed a slightly higher diagnostic value Since these clinicopathological parameters are indica- than AFP. As for miR-101-5p, the SROC was 0.80 tors of tumor deterioration and progression, monitor- (95% CI: 0.76–0.83), a little bit lower than AFP, but it ing the level of miR-101 may have a certain also showed a moderate value for HCC diagnosis, significance in the progression of HCC. which is comparable to AFP’s diagnostic value. Accumulating studies have indicated that dysregula- Even though miR-101-3p/5p expression showed a tion of circulating miRNAs could be a biomarker of high diagnostic value for HCC, the heterogeneity tumorigenesis, development and invasion in various among the studies must be considered. Since our study cancers including prostate cancer, gastric cancer, ovar- indicated that heterogeneity from the threshold effect ian cancer, and lung cancer [30,46–49]. A was absent, we deduced that the heterogeneity may be diagnostic value for circulating miR-101-3p/5p in caused by the different data platforms and the large HCC has also been reported [50,51]. Both of these gaps between each study. Considering that the number studies validated that a lower miR-101-3p/5p level had of studies was small, we did not conduct a subgroup diagnostic potential for HCC. However, due to the analysis.

Fig. 13. The PPI network of miR-101-5p potential targets. Both the color and the size of the nodes reflect the connectivity degrees of two nodes; nodes with a blue color are perceived as hub nodes.

80 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

Subsequently, bioinformatic analysis was performed proliferation and promotes apoptosis by targeting to determine the molecular mechanism of miR-101-3p/ JAK2 [49]. Previous studies reveal that miR-101-3p 5p in HCC. In the past, researchers exploring the might regulate the occurrence and development of HCC molecular mechanism of miRNAs only concentrated on by targeting various genes, and tumorigenesis probably one or two target genes. For example, Varambally et al. results from the abnormality of multiple genes. Of [52] first reported that EZH2 was the target gene of course, the correlation of those potential key genes of miR-101 (3p and 5p were not distinguished) several miR-101-3p needs further experimental validation. years ago. Another study confirmed that miR-101 (3p Next, a functional analysis of these target genes in vitro and 5p were not distinguished) inhibits HCC progres- and in vivo will need to be conducted, such as by RNA sion and through EZH2 down-regulation interference and cellular transfection, luciferase reporter [53]. Liu et al. [54] identified another target gene of miR assay, western blot and so on. miR-101-5p possibly tar- 101 (3p and 5p were not distinguished), VEGF C, which gets ESR1, KRAS, CREBBP, FOXO1 and SMAD3 promotes invasion and migration. MCL-1 and COX-2, through different pathways. Among them, KRAS,a which play a role in tumorigenesis, have also been iden- Kirsten ras oncogene homolog, was reported to have tified as the target genes of miR-101 (3p and 5p were functional synergy with HBx in HCC initiation and not distinguished) [31,55]. In addition, metastasis of progression [62], and FOXO1 has been proposed to HCC has been shown to be affected by different target inhibit EMT transcriptional activators in HCC [63,64]. genes of miR-101 (3p and 5p were not distinguished), Additionally, Hishida et al. [65] indicated that ESR1 is such as STMN1 [56] and PTEN [57]. Since a single a tumor suppressor gene in HCC. Taken together, the miRNA can target multiple genes to achieve its biologi- hub genes identified may perform key roles in HCC. cal and clinical functions, the exploration of the rele- Further investigation appears to be necessary to con- vant gene network can reveal the widespread molecular firm their exact function in HCC. mechanism of miR-101-3p/5p. Hence, we identified Taken together, the present study validated the potential target genes of miR-101-3p/5p in silico. More- down-regulation of the two opposing strands, miR- over, we further narrowed the list by analyzing the 101-3p and miR-101-5p, in HCC clinical specimens; genes that overlapped with the differentially expressed however, miR-101-3p held a greater value for HCC genes of HCC identified via NLP. Next, these target diagnosis. Bioinformatic analysis revealed that miR- genes were subjected to KEGG pathway annotation 101-3p and miR-101-5p are involved in the same or and GO enrichment analysis by using the DAVID. The similar signaling pathways through regulating a differ- target genes of both miR-101-3p and miR-101-5p are ent set of target genes. The fact that miR-101-3p and involved in pathways in cancer, hepatitis B and the miR-101-5p are involved in these signaling pathways MAPK signaling pathway. These results reveal that suggests that the expression of miR-101-3p and miR- miR-101 probably contributes to the tumorigenesis and 101-5p is close in HCC tissues, and they may function metastasis of HCC. Previous studies have reported the cooperatively with each other in the differentiation, role of these pathways in liver cancer [58,59]. The GO proliferation and development of HCC. term analysis indicated that these potential target genes In conclusion, we provide a comprehensive analysis of miR-101-3p/5p were significantly involved in the reg- of miR-101-3p/5p and evaluated the value of miR-101- ulation of the cell cycle and cell proliferation, which are 3p and miR-101-5p as biomarkers for the early diag- associated with tumor occurrence or stepwise develop- nosis of HCC. In addition, we investigated the ment. prospective molecular mechanisms of these two oppos- Furthermore, we constructed the PPI network with ing strands in silico. Our results provide a deeper potential target genes, showing that miR-101-3p proba- understanding of the role of miR-101-3p/5p in HCC bly targets FOS, SMARCA4, MAPK1, GSK3B and and facilitate the possible development of a miRNA- JAK2 to exert its function in HCC. Li et al. [60] based targeted therapy of HCC. However, several limi- reported that FOS acts as a regulator of cell prolifera- tations should be considered in this study. First, the tion, differentiation and transformation, and miR-101 total number of studies included was limited; second, inhibits cell invasion and migration via down-regulation further experiments in vitro and in vivo are still of FOS. MAPK1 has also been reported to be involved required to confirm the function of the target genes. in a variety of cellular processes, such as differentiation, proliferation and development through the MAPK Acknowledgements pathway [61]. JAK2 is a protein tyrosine kinase, and recent evidence has demonstrated that miR-101 (3p and The study was supported by the Fund of Youth 5p were not distinguished) inhibits breast cancer cell Science Foundation of Guangxi Medical University

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 81 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

(GXMUYSF201624). The funder had no role in study 12 Wang N, Chen P, Huang LP and Wang TZ (2016) design, data collection and analysis, decision to pub- Prognostic significance of microRNA-10b lish, or preparation of the manuscript. overexpression in breast cancer: a meta-analysis. Genet Mol Res 15, gmr.15027350. https://doi.org/10.4238/gmr. 15027350 Author contributions 13 Slattery ML, Herrick JS, Pellatt DF, Mullany LE, XY and Y-YP contributed equally in this study. XY, Stevens JR, Wolff E, Hoffman MD, Wolff RK and Y-YP and PL collected data and wrote the paper Samowitz W (2016) Site-specific associations between draft; R-QH and J-MC contributed to the analysis; miRNA expression and survival in colorectal cancer cases. Oncotarget 7, 60193–60205. JM, GC and HY designed the study, supervised the 14 Xu P, Zhu Y, Sun B and Xiao Z (2016) Colorectal report and corrected the manuscript. cancer characterization and therapeutic target prediction based on microRNA expression profile. Sci References Rep 6, 20616. 15 Tang S, Wu WK, Li X, Wong SH, Wong N, Chan 1 Siegel RL, Miller KD and Jemal A (2017) Cancer MT, Sung JJ and Yu J (2016) Stratification of digestive statistics, 2017. CA Cancer J Clin 67,7–30. cancers with different pathological features and survival 2 Sherman M (2010) Hepatocellular carcinoma: outcomes by MicroRNA expression. Sci Rep 6, 24466. epidemiology, surveillance, and diagnosis. Semin Liver 16 Mott JL (2009) MicroRNAs involved in tumor Dis 30,3–16. suppressor and oncogene pathways: implications for 3 Xie Y, Yao Q, Butt AM, Guo J, Tian Z, Bao X, Li H, hepatobiliary neoplasia. Hepatology 50, 630–637. Meng Q and Lu J (2014) Expression profiling of serum 17 Visone R, Petrocca F and Croce CM (2008) Micro- microRNA-101 in HBV-associated chronic hepatitis, RNAs in gastrointestinal and liver disease. liver cirrhosis, and hepatocellular carcinoma. Cancer Gastroenterology 135, 1866–1869. Biol Ther 15, 1248–1255. 18 Almeida MI, Reis RM and Calin GA (2011) 4 Calin GA and Croce CM (2006) MicroRNA MicroRNA history: discovery, recent applications, and signatures in human cancers. Nat Rev Cancer 6, next frontiers. Mutat Res 717,1–8. 857–866. 19 Almeida MI, Nicoloso MS, Zeng L, Ivan C, Spizzo R, 5 Bartel DP (2004) MicroRNAs: genomics, biogenesis, Gafa R, Xiao L, Zhang X, Vannini I, Fanini F et al. mechanism, and function. Cell 116, 281–297. (2012) Strand-specific miR-28-5p and miR-28-3p have 6 Harfe BD (2005) MicroRNAs in vertebrate distinct effects in colorectal cancer cells. development. Curr Opin Genet Dev 15, Gastroenterology 142, 886–896.e9. 410–415. 20 Gong J, Wu Y, Zhang X, Liao Y, Sibanda VL, Liu W 7 Jiang J, Gusev Y, Aderca I, Mettler TA, Nagorney and Guo AY (2014) Comprehensive analysis of human DM, Brackett DJ, Roberts LR and Schmittgen TD small RNA sequencing data provides insights into (2008) Association of microRNA expression in expression profiles and miRNA editing. RNA Biol 11, hepatocellular carcinomas with hepatitis infection, 1375–1385. cirrhosis, and patient survival. Clin Cancer Res 14, 21 Yan F, Shen N, Pang J, Xie D, Deng B, Molina JR, 419–427. Yang P and Liu S (2014) Restoration of miR-101 8 Fu Y, Wei X, Tang C, Li J, Liu R, Shen A and Wu Z suppresses lung tumorigenesis through inhibition of (2013) Circulating microRNA-101 as a potential DNMT3a-dependent DNA methylation. Cell Death Dis biomarker for hepatitis B virus-related hepatocellular 5, e1413. carcinoma. Oncol Lett 6, 1811–1815. 22 Yao YL, Ma J, Wang P, Xue YX, Li Z, Zhao LN, Li 9 Liao Q, Han P, Huang Y, Wu Z, Chen Q, Li S, Ye J ZQ, Feng TD and Liu YH (2015) miR-101 acts as a and Wu X (2015) Potential role of circulating tumor suppressor by targeting Kruppel-like factor 6 in microRNA-21 for hepatocellular carcinoma diagnosis: a glioblastoma stem cells. CNS Neurosci Ther 21,40–51. meta-analysis. PLoS ONE 10, e0130677. 23 Wei X, Tang C, Lu X, Liu R, Zhou M, He D, Zheng 10 Hou C, Tan G and Feng S (2014) Clinical significance D, Sun C and Wu Z (2015) MiR-101 targets DUSP1 to of microRNA expressions in diagnosing uterine cancer regulate the TGF-beta secretion in sorafenib inhibits and predicting lymph node metastasis. Tumour Biol 35, macrophage-induced growth of hepatocarcinoma. 10789–10798. Oncotarget 6, 18389–18405. 11 Wang H, Wu S, Zhao L, Zhao J, Liu J and Wang Z 24 Zhou X, Xia Y, Li L, and Zhang G (2015) MiR-101 (2015) Clinical use of microRNAs as potential non- inhibits cell growth and tumorigenesis of Helicobacter invasive biomarkers for detecting non-small cell lung pylori related gastric cancer by repression of SOCS2. cancer: a meta-analysis. Respirology 20,56–65. Cancer Biol Ther 16, 160–169.

82 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. X. Yang et al. Diagnostic value and bioinformatics analysis of miR-101 in HCC

25 Ye Z, Yin S, Su Z, Bai M, Zhang H, Hei Z and Cai S migration, and invasion in hepatocellular carcinoma by (2016) Downregulation of miR-101 contributes to targeting SET. Onco Targets Ther 9, 3281–3289. epithelial-mesenchymal transition in cisplatin resistance 37 Mourelatos Z, Dostie J, Paushkin S, Sharma A, of NSCLC cells by targeting ROCK2. Oncotarget 7, Charroux B, Abel L, Rappsilber J, Mann M and 37524–37535. Dreyfuss G (2002) miRNPs: a novel class of 26 Hou J, Lin L, Zhou W, Wang Z, Ding G, Dong Q, ribonucleoproteins containing numerous microRNAs. Qin L, Wu X, Zheng Y, Yang Y et al. (2011) Genes Dev 16, 720–728. Identification of miRNomes in human liver and 38 Chen L and Jin H (2014) MicroRNAs as novel hepatocellular carcinoma reveals miR-199a/b-3p as biomarkers in the diagnosis of non-small cell lung therapeutic target for hepatocellular carcinoma. Cancer cancer: a meta-analysis based on 20 studies. Tumour Cell 19, 232–243. Biol 35, 9119–9129. 27 Wei X, Xiang T, Ren G, Tan C, Liu R, Xu X and Wu 39 Higgins JP and Thompson SG (2002) Quantifying Z (2013) miR-101 is down-regulated by the hepatitis B heterogeneity in a meta-analysis. Stat Med 21,1539–1558. virus x protein and induces aberrant DNA methylation 40 Dweep H and Gretz N (2015) miRWalk2.0: a by targeting DNA methyltransferase 3A. Cell Signal 25, comprehensive atlas of microRNA-target interactions. 439–446. Nat Methods 12, 697. 28 Zhang Y, Guo X, Xiong L, Kong X, Xu Y, Liu C, 41 Chang EK, Yu CY, Clarke R, Hackbarth A, Sanders Zou L, Li Z, Zhao J and Lin N (2012) MicroRNA-101 T, Esrailian E, Hommes DW and Runyon BA (2016) suppresses SOX9-dependent tumorigenicity and Defining a patient population with cirrhosis: an promotes favorable prognosis of human hepatocellular automated algorithm with natural language processing. carcinoma. FEBS Lett 586, 4362–4370. J Clin Gastroenterol 50, 889–894. 29 Sheng Y, Li J, Zou C, Wang S, Cao Y, Zhang J, 42 Alsawas M, Alahdab F, Asi N, Li DC, Wang Z and Huang A and Tang H (2014) Downregulation of miR- Murad MH (2016) Natural language processing: use in 101-3p by hepatitis B virus promotes proliferation and EBM and a guide for appraisal. Evid Based Med 21, migration of hepatocellular carcinoma cells by targeting 136–138. Rab5a. Arch Virol 159, 2397–2410. 43 Zhang X, Tang W, Chen G, Ren F, Liang H, Dang Y 30 Wang HJ, Ruan HJ, He XJ, Ma YY, Jiang XT, Xia and Rong M (2016) An encapsulation of gene YJ, Ye ZY and Tao HQ (2010) MicroRNA-101 is signatures for hepatocellular carcinoma, microRNA-132 down-regulated in gastric cancer and involved in cell predicted target genes and the corresponding overlaps. migration and invasion. Eur J Cancer 46, 2295–2303. PLoS ONE 11, e0159498. 31 Su H, Yang JR, Xu T, Huang J, Xu L, Yuan Y and 44 Huang WT, Wang HL, Yang H, Ren FH, Luo YH, Zhuang SM (2009) MicroRNA-101, down-regulated in Huang CQ, Liang YY, Liang HW, Chen G and Dang hepatocellular carcinoma, promotes apoptosis and YW (2016) Lower expressed miR-198 and its potential suppresses tumorigenicity. Cancer Res 69, 1135–1142. targets in hepatocellular carcinoma: a 32 Liu N, Zhang L, Wang Z, Cheng Y, Zhang P, Wang clinicopathological and in silico study. Onco Targets X, Wen W, Yang H, Liu H, Jin W et al. (2017) Ther 9, 5163–5180. MicroRNA-101 inhibits proliferation, migration and 45 Zamora J, Abraira V, Muriel A, Khan K and invasion of human glioblastoma by targeting SOX9. Coomarasamy A (2006) Meta-DiSc: a software for Oncotarget 8, 19244–19254. meta-analysis of test accuracy data. BMC Med Res 33 He S, Hu XW, Wang D, Han LF, Zhang DC and Wei Methodol 6, 31. C (2016) Accuracy of microRNAs for the diagnosis of 46 Wang J, Ye H, Zhang D, Hu Y, Yu X, Wang L, Zuo hepatocellular carcinoma: a systematic review and meta- C, Yu Y, Xu G and Liu S (2016) MicroRNA-410-5p as analysis. Clin Res Hepatol Gastroenterol 40,405–417. a potential serum biomarker for the diagnosis of 34 Ha SY, Kim JH, Yang JW, Bae H, Cho HY and Park prostate cancer. Cancer Cell Int 16, 12. CK (2016) Expression of DBC1 is associated with poor 47 Cochetti G, Poli G, Guelfi G, Boni A, Egidi MG and prognosis in hepatitis virus-related hepatocellular Mearini E (2016) Different levels of serum microRNAs carcinoma. Pathol Res Pract 212, 616–621. in prostate cancer and benign prostatic hyperplasia: 35 Kindrat I, Tryndyak V, de Conti A, Shpyleva S, evaluation of potential diagnostic and prognostic role. Mudalige TK, Kobets T, Erstenyuk AM, Beland FA Onco Targets Ther 9, 7545–7553. and Pogribny IP (2016) MicroRNA-152-mediated 48 Liu L, Guo J, Yu L, Cai J, Gui T, Tang H, Song L, dysregulation of hepatic transferrin receptor 1 in liver Wang J, Han F, Yang C et al. (2014) miR-101 carcinogenesis. Oncotarget 7, 1276–1287. regulates expression of EZH2 and contributes to 36 Jin H, Yu M, Lin Y, Hou B, Wu Z, Li Z and Sun J progression of and cisplatin resistance in epithelial (2016) MiR-502-3P suppresses cell proliferation, . Tumour Biol 35, 12619–12626.

FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 83 Diagnostic value and bioinformatics analysis of miR-101 in HCC X. Yang et al.

49 Wang L, Li L, Guo R, Li X, Lu Y, Guan X, Gitau SC, carcinogenic mechanism of anti-miR-101. Oncol Rep 36, Wang L, Xu C, Yang B et al. (2014) miR-101 promotes 2184–2192. breast cancer cell apoptosis by targeting Janus kinase 2. 58 Li L, Zheng BB, Ma LS, Sun X, Chang JJ, Xie WD Cell Physiol Biochem 34, 413–422. and Li X (2014) Telekin suppresses human 50 Zhuang C, Jiang W, Huang D, Xu L, Yang Q, hepatocellular carcinoma cells in vitro by inducing G2/ Zheng L, Wang X and Hu L (2016) Serum miR-21, M phase arrest via the p38 MAPK signaling pathway. miR-26a and miR-101 as potential biomarkers of Acta Pharmacol Sin 35, 1311–1322. hepatocellular carcinoma. Clin Res Hepatol 59 Yamaguchi F, Hirata Y, Akram H, Kamitori K, Dong Gastroenterol 40, 386–396. Y, Sui L and Tokuda M (2013) FOXO/TXNIP 51 He S, Zhang DC and Wei C (2015) MicroRNAs as pathway is involved in the suppression of hepatocellular biomarkers for hepatocellular carcinoma diagnosis and carcinoma growth by glutamate antagonist MK-801. prognosis. Clin Res Hepatol Gastroenterol 39, 426–434. BMC Cancer 13, 468. 52 Varambally S, Cao Q, Mani RS, Shankar S, Wang X, 60 Li S, Fu H, Wang Y, Tie Y, Xing R, Zhu J, Sun Z, Ateeq B, Laxman B, Cao X, Jing X, Ramnarayanan K Wei L and Zheng X (2009) MicroRNA-101 regulates et al. (2008) Genomic loss of microRNA-101 leads to expression of the v-fos FBJ murine osteosarcoma viral overexpression of histone methyltransferase EZH2 in oncogene homolog (FOS) oncogene in human cancer. Science 322, 1695–1699. hepatocellular carcinoma. Hepatology 49, 1194–1202. 53 Xu L, Beckebaum S, Iacob S, Wu G, Kaiser GM, 61 Delire B and Starkel P (2015) The Ras/MAPK pathway Radtke A, Liu C, Kabar I, Schmidt HH, Zhang X and hepatocarcinoma: pathogenesis and therapeutic et al. (2014) MicroRNA-101 inhibits human implications. Eur J Clin Invest 45, 609–623. hepatocellular carcinoma progression through EZH2 62 Ye H, Zhang C, Wang BJ, Tan XH, Zhang WP, Teng downregulation and increased cytostatic drug Y and Yang X (2014) Synergistic function of Kras sensitivity. J Hepatol 60, 590–598. mutation and HBx in initiation and progression of 54 Liu Z, Wang J, Mao Y, Zou B and Fan X (2016) hepatocellular carcinoma in mice. Oncogene 33, 5133– MicroRNA-101 suppresses migration and invasion via 5138. targeting vascular endothelial growth factor-C in 63 Lee SY, Lee GR, Woo DH, Park NH, Cha HJ, Moon hepatocellular carcinoma cells. Oncol Lett 11, YH and Han IS (2013) Depletion of Aurora A leads to 433–438. upregulation of FoxO1 to induce cell cycle arrest in 55 Strillacci A, Griffoni C, Sansone P, Paterini P, Piazzi hepatocellular carcinoma cells. Cell Cycle 12,67–75. G, Lazzarini G, Spisni E, Pantaleo MA, Biasco G and 64 Dong T, Zhang Y, Chen Y, Liu P, An T, Zhang J, Tomasi V (2009) MiR-101 downregulation is involved Yang H, Zhu W and Yang X (2017) FOXO1 inhibits in cyclooxygenase-2 overexpression in human colon the invasion and metastasis of hepatocellular carcinoma cancer cells. Exp Cell Res 315, 1439–1447. by reversing ZEB2-induced epithelial-mesenchymal 56 Xu Y, An Y, Wang Y, Zhang C, Zhang H, Huang C, transition. Oncotarget 8, 1703–1713. Jiang H, Wang X and Li X (2013) miR-101 inhibits 65 Hishida M, Nomoto S, Inokawa Y, Hayashi M, Kanda autophagy and enhances cisplatin-induced apoptosis in M, Okamura Y, Nishikawa Y, Tanaka C, Kobayashi hepatocellular carcinoma cells. Oncol Rep 29, 2019– D, Yamada S et al. (2013) 1 gene as 2024. a tumor suppressor gene in hepatocellular carcinoma 57 Lv X, Li J and Yang B (2016) Clinical effects of miR- detected by triple-combination array analysis. Int J 101 on prognosis of hepatocellular carcinoma and Oncol 43,88–94.

84 FEBS Open Bio 8 (2018) 64–84 ª 2017 The Authors. Published by FEBS Press and John Wiley & Sons Ltd.